Bug 166122 - After the KB5055523 win11 24H2 update, libreoffice only opens in safe mode
Summary: After the KB5055523 win11 24H2 update, libreoffice only opens in safe mode
Status: NEW
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: LibreOffice (show other bugs)
Version:
(earliest affected)
25.2.2.2 release
Hardware: x86-64 (AMD64) Windows (All)
: medium normal
Assignee: Not Assigned
URL:
Whiteboard:
Keywords: haveBacktrace
: 166228 166776 (view as bug list)
Depends on:
Blocks: Skia
  Show dependency treegraph
 
Reported: 2025-04-10 10:10 UTC by zdziekonski
Modified: 2025-06-13 17:46 UTC (History)
13 users (show)

See Also:
Crash report or crash signature:


Attachments
Win11 System Information (.nfo file) -- see comment #11 (1.83 MB, text/x-nfo)
2025-04-15 03:12 UTC, Jim Avera
Details
Minidump (5.03 MB, application/octet-stream)
2025-04-15 19:05 UTC, Kai-Christian Zellemann
Details
DxDiag output from a Hyper-V-VM (43.47 KB, text/plain)
2025-04-25 11:01 UTC, Kai-Christian Zellemann
Details
DxDiag output from a VMware Horizon-VM (89.81 KB, text/plain)
2025-04-25 11:02 UTC, Kai-Christian Zellemann
Details

Note You need to log in before you can comment on or make changes to this bug.
Description zdziekonski 2025-04-10 10:10:38 UTC
Description:
After the KB5055523 win11 24H2 update, libreoffice only opens in safe mode and after rolling it back everything is ok. This is the record from the event viewer: Faulting application name: soffice.bin, version: 25.2.2.2, timestamp: 0x67ddc0cd
Faulting module name: ntdll.dll, version: 10.0.26100.3775, timestamp: 0x5e4be250
Exception code: 0xc0000374
Fault offset: 0x0000000000115f55
Faulting process ID: 0x1A74
Faulting application start time: 0x1DBA9FC8F527138
Faulting application path: C:\Program Files\LibreOffice\program\soffice.bin

Faulting module path: C:\WINDOWS\SYSTEM32\ntdll.dll
Report Id: 22f448e2-e748-4fd4-ab08-72805d06206f
Faulting full package name:
Faulting package relative application id:


Steps to Reproduce:
1.uninstall KB5055523
2.
3.

Actual Results:
libreoffice works normally

Expected Results:
install KB5055523 and libreoffice works normally


Reproducible: Always


User Profile Reset: No

Additional Info:
[Information automatically included from LibreOffice]
Locale: pl
Module: StartModule
[Information guessed from browser]
OS: Windows (All)
OS is 64bit: no
Comment 1 Mike Kaganski 2025-04-10 10:43:47 UTC
Works fine here with Windows 11 version 24H2 (OS Build 26100.3775) with 2025-04 Cumulative Update for Windows 11 Version 24H2 for x64-based Systems (KB5055523) installed, and using Version: 25.2.2.2 (X86_64) / LibreOffice Community
Build ID: 7370d4be9e3cf6031a51beef54ff3bda878e3fac
CPU threads: 24; OS: Windows 11 X86_64 (10.0 build 26100); UI render: Skia/Vulkan; VCL: win
Locale: de-DE (en_US); UI: en-GB
Calc: CL threaded
Comment 2 Mike Kaganski 2025-04-10 10:46:31 UTC
(In reply to zdziekonski from comment #0)
> ...
> Fault offset: 0x0000000000115f55
> ...
> [Information guessed from browser]
> OS: Windows (All)
> OS is 64bit: no

The "OS is 64bit: no" is obviously wrong; that is definitely a 64-bit OS, so no difference with my working configuration in that regard.
Comment 3 V Stuart Foote 2025-04-10 11:24:27 UTC
Likewise no issue with Win11 24H2 (OS Build 26100.3775) of KB5055523

Version: 25.2.2.2 (X86_64) / LibreOffice Community
Build ID: 7370d4be9e3cf6031a51beef54ff3bda878e3fac
CPU threads: 28; OS: Windows 11 X86_64 (10.0 build 26100); UI render: Skia/Vulkan; VCL: win
Locale: en-US (en_US); UI: en-US
Calc: CL threaded
Comment 4 Kai-Christian Zellemann 2025-04-12 17:34:53 UTC
Unfortunately, I may have the same issue after installing KB5053656. LibreOffice can also only be opened in safe mode or when hardware acceleration is disabled. I can reproduce the behavior on a second, freshly installed machine.

"Fehlerhafter Anwendungsname: soffice.bin, Version: 25.2.2.2, Zeitstempel: 0x67ddc0cd
Fehlerhafter Modulname: ntdll.dll, Version: 10.0.26100.3775, Zeitstempel: 0x5e4be250
Ausnahmecode: 0xc0000374
Fehleroffset: 0x0000000000115f55
Fehlerhafte Prozess-ID: 0x1928
Fehlerhafte Anwendungsstartzeit: 0x1DBABCEAAF333FC
Fehlerhafter Anwendungspfad: C:\Program Files\LibreOffice\program\soffice.bin
Fehlerhafter Modulpfad: C:\Windows\SYSTEM32\ntdll.dll
Berichts-ID: 0abced20-8636-4b35-abc3-3d0b7a894442
Vollständiger Name des fehlerhaften Pakets: 
Fehlerhafte paketbezogene Anwendungs-ID:"

https://bugs.documentfoundation.org/show_bug.cgi?id=166080 probably has the same issue.

Version: 25.2.2.2 (X86_64) / LibreOffice Community
Build ID: 7370d4be9e3cf6031a51beef54ff3bda878e3fac
CPU threads: 16; OS: Windows 11 X86_64 (10.0 build 26100); UI render: default; VCL: win
Locale: de-DE (de_DE); UI: de-DE
Calc: threaded
Comment 5 V Stuart Foote 2025-04-14 11:53:20 UTC
@Mike, * 
Since the LO upgrade install launch defaults are GPU testing with HA set and skia/Vulkan rendering with fall back, might the issue be the skia testing? So both skia/Vulcan and the skia based raster framing failing (the skia fall back also fails). 

Then the default with a Safe mode (which I'd thought was to use skia raster framing, but maybe not) also fails but actually launches using GDI+ 'default; rendering? With or without the Hardware Acceleration check box, which I have always perceived as HA under os/DE control (i.e. more rendering of VCL by CPU, but less extensive than the skia paths).    

Also, still could be a GPU driver (which *can* get clobbered) with Windows build updates. And users rolling back to prior builds have profiles configured to skia/raster or GDI 'default' rendering, and they squeak by on 'default' rendering.

Think we need to see systems specifics of iGPU/dGPU present, and the drivers for each, for folks having this issue with the KB5055523 (OS Build 26100.3775) or earlier KB5053656 (OS Build 26100.3624 )Windows builds applied.

Users can run the Windows 'Device Manager' panel and from the Display adapters identify the GPU and drivers, alternatively the Components -> Display panel of a run of msinfo32.exe

For Vulkan support, Intel, nVidia and AMD all include the 'vulkaninfo' command in their GPU driver bundles that Windows should access. From that long report details in the: Device Properties and Extensions: VkPhysicalDeviceProperties:
for the active GPU would be useful.
Comment 6 Mike Kaganski 2025-04-14 12:07:16 UTC
(In reply to V Stuart Foote from comment #5)

Do you refer to bug 164788 and commit 9735a9377b3013943e6a245b5afef57de738a0f6 ?
I do not see how my change could create a *new* issue here. Of course, Skia could be the problem - but not that change.

IMO, we actually need a minidump.
Comment 7 Mike Kaganski 2025-04-14 12:08:37 UTC
(In reply to Mike Kaganski from comment #6)
> Do you refer to bug 164788 and commit
> 9735a9377b3013943e6a245b5afef57de738a0f6 ?

Ah - obviously no, that is completely unrelated, only appearing in the future 25.8.
Comment 8 V Stuart Foote 2025-04-14 13:18:13 UTC
(In reply to Mike Kaganski from comment #7)
> (In reply to Mike Kaganski from comment #6)
> > Do you refer to bug 164788 and commit
> > 9735a9377b3013943e6a245b5afef57de738a0f6 ?
> 
> Ah - obviously no, that is completely unrelated, only appearing in the
> future 25.8.

And no nothing specific for the GPU render "testing", just that we have a history first with the OpenGL paths, and now with the skia based rendering that driver implementations can interfere with launch.

Operating system updates like these for Windows have induced rollback to less functional default drivers. So that is not necessarily something in our LO rendering that we have any control over.

Think we stil want to test and perform fallback (both graphics rendering and OpenCL support)--that is allow the higher performant devices by default and fall back on testing fails. As opposed to taking least performant (but more stable) device choices by default--and forcing users to optimize for better/best performance.

We don't know why users are reporting failure on an os upgrade, we've not necessarily changed something that would cause it but we could have.

Needinfo from users to tease out what's going on with their os update and impact on GPU / OpenCL support, and does it actually happen with clean installs at 24.8.6 or not (and only at 25.2.2 or later).
Comment 9 V Stuart Foote 2025-04-14 13:44:21 UTC
And I guess we could also ask for /a administrative installs in parallel of a nightly master against 25.8.0 improves things when KB5055523 (OS Build 26100.3775) or earlier KB5053656 (OS Build 26100.3624 )Windows builds are applied with their HW configurations.

https://wiki.documentfoundation.org/Installing_in_parallel

https://dev-builds.libreoffice.org/daily/master/


Then adjust the User Configuration for the parallel installation:

Use a file manager to find the file "bootstrap.ini" in the new instances "program" directory and edit the row from original:

"UserInstallation=$SYSUSERCONFIG/LibreOffice/4"

to a new: 

"UserInstallation=$ORIGIN/../Data/settings"

The $ORIGIN is relative to the installation, so keeps it self contained and isolated from your other instances. While $SYSUSERCONFIG would put it into the os/DE user profile, e.g. %APPDATA%/LibreOfficeDev (normally C:\users\<username>\AppData\Roaming), avoid that if you work with a lot of copies. Better to keep each instance isolated.

To start the application, run "soffice.exe" in the new "program" directory from Explorer or from a shortcut link on desktop.

Note: one should keep old and new versions of bootstrap.ini in a safe place.
Comment 10 V Stuart Foote 2025-04-14 14:14:36 UTC
(In reply to V Stuart Foote from comment #5)
> 
> Think we need to see systems specifics of iGPU/dGPU present, and the drivers
> for each, for folks having this issue with the KB5055523 (OS Build
> 26100.3775) or earlier KB5053656 (OS Build 26100.3624 )Windows builds
> applied.
> 

For folks on Win11 having this issue, please post the following additional items, run from a command window:

1. 'winver' details for your os/DE, we need the 26100.xxxx number, like this:

Microsoft Windows
Version 24H2 (OS Build 26100.3775)

2. GPU device properties from a run of 'vulkaninfo | more' partway down the report, like this:

Device Properties and Extensions:
=================================
GPU0:
VkPhysicalDeviceProperties:
---------------------------
        apiVersion        = 1.4.303 (4210991)
        driverVersion     = 572.60.0.0 (2400124928)
        vendorID          = 0x10de
        deviceID          = 0x2803
        deviceType        = PHYSICAL_DEVICE_TYPE_DISCRETE_GPU
        deviceName        = NVIDIA GeForce RTX 4060 Ti
        pipelineCacheUUID = a552de6d-1faf-8ad5-3314-6af41289f0a5

And also please zip together and add as attachment, the libreoffice OpenCL_devices.log and the skia.log from your "%APPDATA%\LibreOffice\4\cache" directory (normally C:\users\<username>\AppData\Roaming\LibreOffice\4\cache).
Comment 11 Jim Avera 2025-04-15 03:11:09 UTC
Same problem Windows 11 running in a Virtualbox VM.  The graphics is virtualized of course, probably exposing the host's graphics in some way -- which uses built-in Intel graphics (CoffeeLake-S GT2 [UHD Graphics 630]).   Not NVIDIA.

winver result: Version 24H2 (OS BUild 26100.3775)

vulkaninfo -- no such command (but as mentioned, I'm not using NVIDIA)

*** I will attach a .nfo file resulting from File->Save in the Windows System Information app.  It's in xml.  If you load this onto a win11 box and double-click it, the System Info app will show it's contents (has zillions of details about processor, graphics, etc.)


In a command window I tried this:

chdir C:\Program Files\LibreOffice\program
.\soffice.exe 

soffice.exe and soffice.bin briefly appeared in Task Manager, then were quickly replaced by "Windows Error Reporting" (which disappeared a few seconds later).

I know little about Windows and could not find where the error data was stored.

%LOCALAPPDATA%\CrashDumps does not exist
C:\Windows\Minidump does not exist
C:\Windows\Memory.dmp does not exist

---> DOES ANYONE KNOW WHERE TO FIND THE CRASH DATA?

FWIW I also tried running soffice.com and got this error:

.\soffice.com
vkCreateInstance failed
Comment 12 Jim Avera 2025-04-15 03:12:21 UTC
Created attachment 200343 [details]
Win11 System Information (.nfo file) -- see comment #11
Comment 13 Jim Avera 2025-04-15 03:22:54 UTC
Ah, found the Event Viewer which shows this:

  AppName soffice.bin 
  AppVersion 25.2.2.2 
  AppTimeStamp 67ddc0cd 
  ModuleName ntdll.dll 
  ModuleVersion 10.0.26100.3775 
  ModuleTimeStamp 5e4be250 
  ExceptionCode c0000374 
  FaultingOffset 0000000000115f55 
  ProcessId 0x2618 
  ProcessCreationTime 0x1dbadb18a35fffe 
  AppPath C:\Program Files\LibreOffice\program\soffice.bin 
  ModulePath C:\WINDOWS\SYSTEM32\ntdll.dll 
  IntegratorReportId d4ed657a-bba6-4c84-aa56-be391b930d4b 
  PackageFullName  
  PackageRelativeAppId  

And just to be sure we're looking at the same exact code:

SHA1 hash of .\soffice.exe: 3f76dd21f40015887f9a931fe00f016a60c2a0d7

SHA1 hash of .\soffice.bin: 26d92ea2d29ea04feb54b8e5f06c311c8d71f065
Comment 14 Jim Avera 2025-04-15 03:25:39 UTC
Code 0xC0000374 is STATUS_HEAP_CORRUPTION
Comment 15 Jim Avera 2025-04-15 03:41:52 UTC
LO 24.8.6.2 does not have the problem, so maybe a bisect could be done to pinpoint the commit which provokes the problem.

NOT WORKING: 25.2 master

WORKING: Version: 24.8.6.2 (X86_64) / LibreOffice Community
Build ID: 6d98ba145e9a8a39fc57bcc76981d1fb1316c60c
CPU threads: 6; OS: Windows 11 X86_64 (10.0 build 26100); UI render: Skia/Raster; VCL: win
Locale: en-US (en_US); UI: en-US
Calc: threaded
Comment 16 Jim Avera 2025-04-15 04:41:36 UTC
25.2.0.0.alpha1 (the earliest 25.x release) DOES have the problem.

So whatever is causing this was introduced between 24.8.6.2 and 25.2.0.0.alpha1
Comment 17 V Stuart Foote 2025-04-15 13:46:57 UTC
(In reply to Jim Avera from comment #16)
> 25.2.0.0.alpha1 (the earliest 25.x release) DOES have the problem.
> 
> So whatever is causing this was introduced between 24.8.6.2 and
> 25.2.0.0.alpha1

Guess that's possible. 

If I'm reading cgit correctly, the LO 24.8 release is entirely on the skia m116 build. While for the LO 25.2 release skia has been updated to the m130 build. While builds of master against the 25.8 release have just been updated to skia m136 build. Accompanied by LO patching adjustments at each.

The way in which LO Graphics test failures and rendering fallback proceeds could certainly be an issue. In your VM case, would depend on the VirtualBox VMSVG graphics emulation, even on a supported GPU. I'm not sure the UHD Graphics 630 with igx Intel drivers installed to the VM should be considered supported--and probably be LO denylisted to bypass the graphics testing.

Otherwise, the vulkaninfo from the VulkanSDK is supported by all GPU vendors. That it does not show for you in a VirtualBox VM (the VMSVG uses generic drivers Oracle provides) suggesting there is limited Vulkan support provided to the VM, and likely limited emulated igx/skia rendering support for the VM. 

You might check the VM's Display -> Screen -> Extended Features: -> Enable 3D Acceleration check box to see how that behaves.
Comment 18 Kai-Christian Zellemann 2025-04-15 19:05:00 UTC
Created attachment 200352 [details]
Minidump

Environment:
 - Hyper-V VM
 - Windows 11 24H2 Education x64 (Build 26100.3775)
 - Language: German

Only the file "%APPDATA%\LibreOffice\4\user\extensions\buildid" (7370d4be9e3cf6031a51beef54ff3bda878e3fac) exists under "%APPDATA%\LibreOffice\"

vulkaninfo doesn't exist on the system.

Event Viewer entry for the crash dump:

Fehlerhafter Anwendungsname: soffice.bin, Version: 25.2.2.2, Zeitstempel: 0x67ddc0cd
Fehlerhafter Modulname: ntdll.dll, Version: 10.0.26100.3775, Zeitstempel: 0x5e4be250
Ausnahmecode: 0xc0000374
Fehleroffset: 0x0000000000115f55
Fehlerhafte Prozess-ID: 0x24B0
Fehlerhafte Anwendungsstartzeit: 0x1DBAE370D3F0751
Fehlerhafter Anwendungspfad: C:\Program Files\LibreOffice\program\soffice.bin
Fehlerhafter Modulpfad: C:\Windows\SYSTEM32\ntdll.dll
Berichts-ID: c686fe7f-3e48-49c3-a7d0-353138903f2f
Vollständiger Name des fehlerhaften Pakets: 
Fehlerhafte paketbezogene Anwendungs-ID:
Comment 19 Mike Kaganski 2025-04-15 19:41:28 UTC
Definitely Vulkan-related.

> skialo.dll!sk_gpu_test::FreeVulkanFeaturesStructs(const VkPhysicalDeviceFeatures2 * features) Line 886	C++
> skialo.dll!skwindow::internal::VulkanWindowContext::initializeContext() Line 83	C++
> skialo.dll!skwindow::internal::VulkanWindowContext::VulkanWindowContext(const skwindow::DisplayParams & params, std::function<VkSurfaceKHR_T *(VkInstance_T *)> createVkSurface, std::function<bool (VkInstance_T *, VkPhysicalDevice_T *, unsigned int)> canPresent, void(*)()(*)(VkInstance_T *, const char *) instProc) Line 59	C++
> skialo.dll!skwindow::MakeVulkanForWin(HWND__ * hwnd, const skwindow::DisplayParams & params) Line 72	C++
> vclplug_winlo.dll!`anonymous namespace'::createVulkanWindowContext(bool __formal) Line 571	C++
> [Inline Frame] mergedlo.dll!SkiaHelper::getTemporaryWindowContext() Line 546	C++
> mergedlo.dll!SkiaHelper::checkDeviceDenylisted(bool blockDisable) Line 253	C++
> mergedlo.dll!SkiaHelper::isVCLSkiaEnabled() Line 408	C++
> vclplug_winlo.dll!WinSalGraphics::WinSalGraphics(WinSalGraphics::Type eType, bool bScreen, HWND__ * hWnd, SalGeometryProvider * pProvider) Line 636	C++
> vclplug_winlo.dll!WinSalFrame::AcquireGraphics() Line 1063	C++
> mergedlo.dll!vcl::WindowOutputDevice::AcquireGraphics() Line 838	C++
> mergedlo.dll!vcl::Window::ImplInit(vcl::Window * pParent, __int64 nStyle, SystemParentData * pSystemParentData) Line 1100	C++
> mergedlo.dll!ImplBorderWindow::ImplInit(vcl::Window * pParent, __int64 nStyle, BorderWindowStyle nTypeStyle, SystemParentData * pSystemParentData) Line 1566	C++
> [Inline Frame] mergedlo.dll!ImplBorderWindow::{ctor}(vcl::Window *) Line 1594	C++
> [Inline Frame] mergedlo.dll!VclPtrInstance<ImplBorderWindow>::{ctor}(vcl::Window * &) Line 280	C++
> mergedlo.dll!WorkWindow::ImplInit(vcl::Window * pParent, __int64 nStyle, SystemParentData * pSystemParentData) Line 51	C++
> [Inline Frame] mergedlo.dll!IntroWindow::{ctor}() Line 35	C++
> mergedlo.dll!`anonymous namespace'::SplashScreenWindow::SplashScreenWindow(`anonymous-namespace'::SplashScreen * pSplash) Line 125	C++
> [Inline Frame] mergedlo.dll!VclPtr<`anonymous namespace'::SplashScreenWindow>::Create(`anonymous-namespace'::SplashScreen * &&) Line 129	C++
> [Inline Frame] mergedlo.dll!`anonymous-namespace'::SplashScreen::{ctor}() Line 145	C++
> mergedlo.dll!desktop_SplashScreen_get_implementation(com::sun::star::uno::XComponentContext * __formal, const com::sun::star::uno::Sequence<com::sun::star::uno::Any> & __formal) Line 616	C++
> [Inline Frame] cppuhelper3MSC.dll!std::_Func_class<com::sun::star::uno::XInterface *,com::sun::star::uno::XComponentContext *,com::sun::star::uno::Sequence<com::sun::star::uno::Any> const &>::operator()(com::sun::star::uno::XComponentContext * <_Args_0>, const com::sun::star::uno::Sequence<com::sun::star::uno::Any> &) Line 869	C++
> cppuhelper3MSC.dll!cppuhelper::ServiceManager::Data::Implementation::doCreateInstanceWithArguments(const com::sun::star::uno::Reference<com::sun::star::uno::XComponentContext> & context, const com::sun::star::uno::Sequence<com::sun::star::uno::Any> & arguments) Line 717	C++
> cppuhelper3MSC.dll!cppuhelper::ServiceManager::Data::Implementation::createInstanceWithArguments(const com::sun::star::uno::Reference<com::sun::star::uno::XComponentContext> & context, bool singletonRequest, const com::sun::star::uno::Sequence<com::sun::star::uno::Any> & arguments) Line 689	C++
> cppuhelper3MSC.dll!cppuhelper::ServiceManager::createInstanceWithArgumentsAndContext(const rtl::OUString & ServiceSpecifier, const com::sun::star::uno::Sequence<com::sun::star::uno::Any> & Arguments, const com::sun::star::uno::Reference<com::sun::star::uno::XComponentContext> & Context) Line 1013	C++
> mergedlo.dll!desktop::Desktop::OpenSplashScreen() Line 2557	C++
> mergedlo.dll!desktop::Desktop::Main() Line 1356	C++
> mergedlo.dll!ImplSVMain() Line 228	C++
> [Inline Frame] mergedlo.dll!SVMain() Line 246	C++
> mergedlo.dll!soffice_main() Line 121	C++
> [Inline Frame] soffice.bin!sal_main() Line 51	C
> soffice.bin!main(int argc, char * * argv) Line 49	C
Comment 20 nalayna 2025-04-17 09:38:04 UTC Comment hidden (obsolete)
Comment 21 nalayna 2025-04-17 09:39:26 UTC
Hey, I´m new to bugzilla, but I had the same issue, so I am trying to contribute my info here. If you need more details, please contact me. Starting LibreOffice in safe mode and disabling hardware acceleration fixed the problem in 25.2.2.2 for me.

Both these versions had the problem. I tried a new installation first.
LibreOffice Version 25.2.0.3 (LibreOffice_25.2.0_Win_x86-64.msi, download and installation on 12.02.2025, source libreoffice.org) and 25.2.2.2 (LibreOffice_25.2.2_Win_x86-64.msi from libreoffice.org)

I´m running Windows 11 24H2 (Build 26100.3775) with the latest Updates (KB5054979 & KB5055523, installed on 09.04.2025) with a 64-bit installation.
(In reply to V Stuart Foote from comment #10)

> 2. GPU device properties from a run of 'vulkaninfo | more' partway down the
> report, like this:

 the command is failing (even when I run it as admin). Is there another way I can get you the info you need?
Comment 22 V Stuart Foote 2025-04-17 10:52:33 UTC
(In reply to nalayna from comment #20)
Thanks.

> > 2. GPU device properties from a run of 'vulkaninfo | more' partway down the
> > report, like this:
> 
>  the command is failing (even when I run it as admin). Is there another way
> I can get you the info you need?

On Windows builds, if vulkaninfo isn't present, your GPU drivers are likely suspect as is the ability of system to support the skia graphics library calls used for testing Vulkan rendering. A driver reinstall/update from the GPU vendor may be needed, or possibly an install of Vulkan Runtime from the project VulkanSDK site hosted by the LunaRG folks [1].

Otherwise what we need are details of the integrated graphics (iGPU on CPU) GPU and/or discrete graphics (dGPU on PCIe bus), and the device drivers for each

A run of the command 'msinfo32' will open a panel of system details.
From the 'System Summary' panel:
   the top 12 or so lines, showing CPU

From the 'Components' -> 'Display' panel:
   the full set for each GPU registered with the system, that will include the 
   current device and driver sets (but not their Vulkan support directly).

=-ref-=
[1] https://vulkan.lunarg.com/
Comment 23 Mike Kaganski 2025-04-17 10:58:16 UTC
Noel: the stack in comment 19 (from attachment 200352 [details]) makes me suspect the skia version upgrade that happened lately. I have no idea how to repro, and what to do - so just adding you to CC, in case it happens to ring a bell for you?
Comment 24 V Stuart Foote 2025-04-17 11:09:12 UTC
(In reply to V Stuart Foote from comment #22)

> =-ref-=
> [1] https://vulkan.lunarg.com/

Should note that the drivers from the GPU vendor normally include the Vulkan runtime appropriate for the GPU, installing the runtime in addition to the vendor provided drivers may conflict. So a removal and reinstall of the vendor drivers could still be needed.  So a good idea to track down the drivers for your system from either the PC manufacture, or directly from the GPU vendor.

AMD: https://www.amd.com/en/support/download/drivers.html
in the GUI, perform a clean installation

Intel: https://www.intel.com/content/www/us/en/support/articles/000090440/graphics.html
https://answers.microsoft.com/en-us/windows/forum/all/updating-intel-graphics-driver-not-working/3520d477-1e30-4c98-8a27-95c3132eada6

nVidia: https://www.nvidia.com/en-us/drivers/
in the GUI, perform a clean installation
Comment 25 V Stuart Foote 2025-04-17 18:42:48 UTC
*** Bug 166228 has been marked as a duplicate of this bug. ***
Comment 26 Volga 2025-04-19 06:43:37 UTC
(In reply to Jim Avera from comment #11)
> ---> DOES ANYONE KNOW WHERE TO FIND THE CRASH DATA?

You can try to visit C:\Users\[Your user name]\AppData\Local\CrashDumps to get crash dumps.
Comment 27 Kai-Christian Zellemann 2025-04-25 11:01:57 UTC
Created attachment 200512 [details]
DxDiag output from a Hyper-V-VM

The drivers for the virtual graphics cards "Microsoft Hyper-V-Video" and "VMware SVGA 3D/VMware Horizon Indirect Display Driver" do not seem to support Vulkan. When I install the Vulkan Runtime manually, I get the following error message:

"vulkaninfo
WARNING: [Loader Message] Code 0 : windows_read_data_files_in_registry: Registry lookup failed to get layer manifest files.
ERROR: [Loader Message] Code 0 : windows_read_data_files_in_registry: Registry lookup failed to get ICD manifest files.  Possibly missing Vulkan driver?
ERROR: [Loader Message] Code 0 : vkCreateInstance: Found no drivers!
Cannot create Vulkan instance.
This problem is often caused by a faulty installation of the Vulkan driver or attempting to use a GPU that does not support Vulkan.
ERROR at C:\Users\lunarg\work\LunarG\VulkanTests\build\Khronos-Tools\repo\vulkaninfo\vulkaninfo.h:574:vkCreateInstance failed with ERROR_INCOMPATIBLE_DRIVER"
Comment 28 Kai-Christian Zellemann 2025-04-25 11:02:26 UTC
Created attachment 200513 [details]
DxDiag output from a VMware Horizon-VM
Comment 29 Buovjaga 2025-04-30 12:08:42 UTC
For the ones affected, do you also see the issue, if you install an unstable build, Win-x86_64@tb77-TDF from https://dev-builds.libreoffice.org/daily/master/current.html ? The unstable version will not affect your stable user profile. In the unstable version there have been even more Skia library upgrades.
Comment 30 V Stuart Foote 2025-04-30 14:19:46 UTC
(In reply to Buovjaga from comment #29)
> For the ones affected, do you also see the issue, if you install an unstable
> build, Win-x86_64@tb77-TDF from
> https://dev-builds.libreoffice.org/daily/master/current.html ? The unstable
> version will not affect your stable user profile. In the unstable version
> there have been even more Skia library upgrades.

Yes, see the instructions for installing in parallel, and tweak the User profile as indicated (bootstrap.ini edit of the stanza UserInstallation=$ORIGIN/..) to isolate each instance:

https://wiki.documentfoundation.org/Installing_in_parallel
Comment 31 Kai-Christian Zellemann 2025-04-30 17:17:13 UTC
(In reply to Buovjaga from comment #29)
> For the ones affected, do you also see the issue, if you install an unstable
> build, Win-x86_64@tb77-TDF from
> https://dev-builds.libreoffice.org/daily/master/current.html ? The unstable
> version will not affect your stable user profile. In the unstable version
> there have been even more Skia library upgrades.

With this version, the programs can be started, but only on the second launch. On the first launch, a file named "registrymodifications.xcu" is created under "%AppData%\LibreOfficeDev\4\user", and the program immediately exits without displaying a GUI.

registrymodifications.xcu:

<?xml version="1.0" encoding="UTF-8"?>
<oor:items xmlns:oor="http://openoffice.org/2001/registry" xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<item oor:path="/org.openoffice.Office.Common/VCL"><prop oor:name="ForceSkiaRaster" oor:op="fuse"><value>true</value></prop></item>
<item oor:path="/org.openoffice.Setup/L10N"><prop oor:name="ooLocale" oor:op="fuse"><value>de</value></prop></item>
</oor:items>

---
Version: 25.8.0.0.alpha0+ (X86_64) / LibreOffice Community
Build ID: f98d81f008e11fd2786597d82126a6287353c714
CPU threads: 16; OS: Windows 11 X86_64 (build 26100); UI render: Skia/Raster; VCL: win
Locale: de-DE (de_DE); UI: de-DE
Calc: threaded
Comment 32 V Stuart Foote 2025-04-30 23:30:05 UTC
(In reply to Kai-Christian Zellemann from comment #31)
> 
> With this version, the programs can be started, but only on the second
> launch. On the first launch, a file named "registrymodifications.xcu" is
> created under "%AppData%\LibreOfficeDev\4\user", and the program immediately
> exits without displaying a GUI.
> 

OK, that is correct behavior then. Absent a deny list entry for GPU device the Vulkan GPU testing occurs. On failure it should fall back to skia raster based rendering and restart with that mode activated.

For really unsupported GPUs, manual disable of the skia lib based rendering and fall back to GDI+ "default" rendering might still be needed to achieve a functional GUI.
Comment 33 Buovjaga 2025-05-02 11:07:16 UTC
I repro with 25.2.3 in a virtual machine with the same message that was mentioned earlier: vkCreateInstance failed: -9
Comment 34 Buovjaga 2025-05-02 12:18:35 UTC
(In reply to Buovjaga from comment #33)
> I repro with 25.2.3 in a virtual machine with the same message that was
> mentioned earlier: vkCreateInstance failed: -9

Hmm, after installing update KB5055627 just now I can launch 25.2.3. What about the other affected users?
Comment 35 Buovjaga 2025-05-05 14:23:00 UTC
(In reply to Buovjaga from comment #34)
> (In reply to Buovjaga from comment #33)
> > I repro with 25.2.3 in a virtual machine with the same message that was
> > mentioned earlier: vkCreateInstance failed: -9
> 
> Hmm, after installing update KB5055627 just now I can launch 25.2.3. What
> about the other affected users?

Adding a couple of affected users who were not in Cc.
Comment 36 Buovjaga 2025-05-09 10:16:46 UTC
(In reply to Buovjaga from comment #34)
> (In reply to Buovjaga from comment #33)
> > I repro with 25.2.3 in a virtual machine with the same message that was
> > mentioned earlier: vkCreateInstance failed: -9
> 
> Hmm, after installing update KB5055627 just now I can launch 25.2.3. What
> about the other affected users?

Eh, now I notice the issue is back. I needed to launch it three times to get it to run.
Comment 37 Volga 2025-05-10 09:32:11 UTC
What if you have installed this package before you open LibreOffice?

https://apps.microsoft.com/detail/9nqpsl29bfff
Comment 38 V Stuart Foote 2025-05-10 13:15:58 UTC
(In reply to Volga from comment #37)
> What if you have installed this package before you open LibreOffice?
> 
> https://apps.microsoft.com/detail/9nqpsl29bfff

Please notice that utility has not been updated by Microsoft since 2020, and issues folks have reported have been with updates within Win11 handling of skia lib driven testing of Vulkan accelerated rendering.

LibreOffice testing of Vulkan should fail gracefully with fall back to skia raster framing--that is no longer happening. Could be Windows change, but could also be our skia lib updates/patches. Additionally we should probably be blocking more GPU categories from even being tested.

Adding five year old package of compatibility drivers for non-supported graphics devices would not be my first choice for reliable operation.
Comment 39 Volga 2025-05-18 23:22:22 UTC
Sorry Stuart, I've not used Win11 at the moment, but anyway, thank you.
Comment 40 Buovjaga 2025-05-19 17:09:57 UTC
The patch for bug 164788 that forces software rendering upon launch, if a problem is detected, was backported just now to 25.2 and will be available in 25.2.4 due in early June: https://wiki.documentfoundation.org/ReleasePlan/25.2#25.2.4_release
Comment 41 V Stuart Foote 2025-05-29 16:11:22 UTC
*** Bug 166776 has been marked as a duplicate of this bug. ***
Comment 42 V Stuart Foote 2025-06-07 19:18:31 UTC
For what it is worth, just watched a Windows 11 (23H2) update to Windows 11 (24H2), and sure enough the first run of Windows updates rolled back the WDDM 3.1 display driver to a WDDM 2.7 driver.  

Intel Iris Graphics started from 23H2 with the Intel recommended 31.0.101.2135 WDDM 3.1 compliant driver and was rolled back to the Windows Update provided 27.20.100.8853 WDDM 2.7 driver.

Users having issues with the skia/Vulkan rendering should check update of their GPU drivers.

Intel posts a KB item with a workaround to prevent roll back:

https://www.intel.com/content/www/us/en/support/articles/000087834/graphics.html
Comment 43 Buovjaga 2025-06-08 15:23:28 UTC
I installed 25.2.4 and it launches upon first try :)

Version: 25.2.4.3 (X86_64) / LibreOffice Community
Build ID: 33e196637044ead23f5c3226cde09b47731f7e27
CPU threads: 2; OS: Windows 11 X86_64 (10.0 build 26100); UI render: Skia/Raster; VCL: win
Locale: en-US (en_US); UI: en-US
Calc: threaded
Comment 44 V Stuart Foote 2025-06-08 18:05:31 UTC
(In reply to Buovjaga from comment #43)
> I installed 25.2.4 and it launches upon first try :)
> 
> Version: 25.2.4.3 (X86_64) / LibreOffice Community
> Build ID: 33e196637044ead23f5c3226cde09b47731f7e27
> CPU threads: 2; OS: Windows 11 X86_64 (10.0 build 26100); UI render:
> Skia/Raster; VCL: win
> Locale: en-US (en_US); UI: en-US
> Calc: threaded

Was that a clean profile, so skia/Vulkan was tested and fallback. Or residual with skia/raster rendering already set?

Also what GPU and driver combo?
Comment 45 Buovjaga 2025-06-08 18:30:59 UTC
(In reply to V Stuart Foote from comment #44)
> (In reply to Buovjaga from comment #43)
> > I installed 25.2.4 and it launches upon first try :)
> > 
> > Version: 25.2.4.3 (X86_64) / LibreOffice Community
> > Build ID: 33e196637044ead23f5c3226cde09b47731f7e27
> > CPU threads: 2; OS: Windows 11 X86_64 (10.0 build 26100); UI render:
> > Skia/Raster; VCL: win
> > Locale: en-US (en_US); UI: en-US
> > Calc: threaded
> 
> Was that a clean profile, so skia/Vulkan was tested and fallback. Or
> residual with skia/raster rendering already set?
> 
> Also what GPU and driver combo?

Virtual machine. Same with clean profile after shutdown. Earlier it required to launch attempts as mentioned in a previous comment.
Comment 46 V Stuart Foote 2025-06-08 18:54:32 UTC
(In reply to Buovjaga from comment #45)
> > Also what GPU and driver combo?
> 
> Virtual machine. Same with clean profile after shutdown. Earlier it required
> to launch attempts as mentioned in a previous comment.

Cool, means Mike's patch to fallback to skia/raster software rendering is in place for folks since would expect the VMSVGA drivers have no support/inclusion of Vulkan at all.